[time-nuts] Isolated +20V for lamp supply

[Magnus Danielson]

Hi,

On 2020-08-14 11:31, Poul-Henning Kamp wrote:
> --------
> timeok at timeok.it writes:
>>    Corby,
>>
>>    so why did HP use a coaxial cable for the lamp power? it seems nonsense.
> Because it generates a LOT of noise, being in essense a VHF/UHF generator.
>
Recall it is about 100 MHz and 5 W or thereabouts. So, you have the DC
current and you have the 100 MHz AC current. Both can cause interference.

The total isolation approach really only work for DC, at 100 MHz you
jump gaps through capacitive or magnetic coupling, which exists because
of voltage difference or current flow respectively.

The best way to handle both of these is to provide a low impedance path
on the preferred way, as that will dominated the current flow. For DC,
this means adding a low resistive path for the current. For AC it is all
about keeping a low-inductive path and keep the loop area low, which is
also about low-inductance.

The chassi path for ground should be a very low-resistive and
low-inductive path, but the +20 V does not match that. The AC-return
path can be made shorter by providing a capacitive path back, with
inductance on the 20 V feed. Care should be made to ensure inductance
can handle the current through it, as you do not want the core to be
saturated.

If one wants to separate coupling, one can do "star-ground" feed, but
for the 5065 it's mostly the +20V and not the ground that needs it
because of significant difference in conductance area between wires and
chassi, so reasonable to assume wired to provide higher resistance and
thus higher problem with common path. So sure, pulling a separate wire
for the +20V from the PSU to the lamp would help. Less so for the
ground, but you can do that too if you want to, but that is on the
diminishing returns effort list.

In the end, adding more wires in parallel to existing conductions paths
will lower the resistance, and thus lower the cross-coupling. You have a
number of different ways around it. However, for RF-signals you always
want an AC return-path for that current very close to the signal, and if
that happens to bridge between two different DC points you will lower
the voltage difference between these, if that current is high, add a
parallel thick wire. This is also true for AC-power at 50-60 Hz. Many
attempts to break the connection for DC or 50/60 Hz tends to cause
severe isolation problems for the signal, so instead shunting with a
thick wire to connect things tighter together ends up making sense.

So, isolation can maybe be a local strategy, but it doesn't work as one
builds larger things in a box or even a system of boxes. As soon as you
start interconnect you get into issues.

Cheers,
Magnus